Higgs boson signal and background in MCFM

Transcription

1 Higgs boson signal and background in MCFM Zurich, January 11, 2012 Keith Ellis, Fermilab John Campbell, RKE and Ciaran Williams arxiv: , [hep-ph]

2 MCFM MCFM is a unified approach to NLO corrections, both to cross sections and differential distributions: (v6.1, October 2011) Publically available code. J. M. Campbell, R. K. Ellis, C. Williams (main authors) R. Frederix, F. Maltoni, F. Tramontano, S. Willenbrock, G. Zanderighi... Standard Model processes for diboson pairs, vector boson+jets, heavy quarks, Higgs... Decays of unstable particles are included, maintaining spin correlations. Amplitudes (especially the one-loop contributions) calculated ab initio or taken from the literature. 2

3 Vector boson cross sections 3

4 Higgs cross sections Higgs cross sections in MCFM Many of these cross sections are known at NNLO, so MCFM is not state of the art in this regard. The most precise theoretical cross sections were important in the limit-setting phase, and will be important in the coupling measurement phase. Discovery phase? 4

5 H WW production Sensitive mode in the middle mass range 130<mH<2 Mz Background depends on the number of jets. Irreducible background: Standard model WW production. 95% CL Limit on / SM ATLAS Observed Expected ± 1 ± 2 (*) H WW l l -1 Ldt = 2.05 fb s = 7 TeV m H [GeV] 5

6 WW production in MCFM Includes both doubly resonant and singly resonant diagrams with Z/γ*. Full NLO-Virtual corrections from DKS, (hep-ph/ ) Includes gg fermion loop contributions, that are formally higher order, using compact analytic formulae. Cut dependent 6

7 gg WW Dicus, Kao, Repko (87); van der Bij, Glover (89); Binoth et al. (05) no net contribution (Landau-Yang,Furry and EW gauge inv.) straightforward for massless quark loops Contribution of 3rd generation? Diagrams with 1,2 and 3 top quarks. b t We obtain analytic results for mt 0, mb=0. 7

8 Analytic calculation Use unitarity cuts to determine coefficients of basis integrals in general expansion: Many simplifications: all but one bubble coefficient can be obtained from massless result. rational term is not amenable to 4-dim. cut techniques, but is identical to the massless case. Result is not compact enough to write down, but short enough to be evaluated efficiently. 8

9 Impact of the 3rd generation (3/2) 2 Transition from massless result for 3 generations to that for 2 generations. 10% deviation from 2 generation result for actual top mass. 9

10 Full NLO results for WW production Impact of gluongluon contribution enhanced by the Higgs search cuts (e.g. M(ll)<50 GeV). 10

11 Reassessment Signal and background all have the same external particles should compute all, including interference. (d,s) (u,c) b t Is the interference important? At least need to check it s not a bigger effect than quoted NNLO uncertainty. Dittmaier et al. (11) - LHC Higgs cross section WG How do we define signal and background? 11

12 Notation Background only: Signal only: The above is the conventional approach. Include effect of interference: Cross section in the presence of the Higgs, i.e. including also the interference: Compare results for σh and σh,i. 12

13 LHC results Significant difference, 3rd gen. contribution important. 13

14 Interference in the 115<mH<150 GeV (no mt cut). Blow-up of low Higgs mass region. Shows correction to be applied to low mass cross section to account for interference with background process 14

15 Analyzing the interference Separate interference by Re and Im parts of propagator: For a small Higgs width (i.e. light Higgs) the second term is negligible. If the full s-dependence of the first term can be represented by the factor from the propagator, it should vanish on integration (odd about the Higgs mass). this is the case for a similar study of interference in the H γγ channel, no big effects there. Dixon, Siu (03) s-dependence more complicated here because the box diagrams favour large invariant masses (W pairs). 15

16 Visualizing the interference For light masses, the real term is solely responsible for the interference. A combination for high Higgs masses (i.e. widths). Long destructive tail required by unitarity. van der Bij, Glover (89) 16

17 Interference vs. transverse mass Require MT < MH (transverse mass less than putative Higgs mass). Negative value of interference for MT > MH, reflection of role of Higgs in taming high energy behaviour. Thanks to the miracle of Quantum mechanics, perform cut and increase signal and decrease background! 17

18 Effect of interference on Δϕ distribution. Cut-based analyses impose an upper limit on MT this reduces the effect of the interference. ATLAS cuts include 0.75MH <MT<MH 18

19 Effect of interference on Δϕ distribution. Other analysis methods may not impose this cut. Without the MT cut the shapes of distributions can change. 19

20 H -> ZZ* -> e - e + e - e +, e - e + μ - μ +, μ - μ + μ - μ + 20

21 Backgrounds to H -> ZZ-> l + l - l + l - Irreducible background from SM (Z/γ*) boson process. MCFM v6.1 contains both the doubly and singly resonant processes processes + virtual and real QCD corrections. M4l > 2 Mz, t-channel exchange diagrams dominate. M4l < 120GeV, influence of singly resonant diagrams. 21

22 Identical particle interference, e.g. μ - μ + μ - μ + Effect of identical leptons in the final state, pp e - e + e - e +, μ - μ + μ - μ + Identical particle interference decreases cross section in the region M4l>110GeV In this region the effect is smallish < 5%, (cf. Melia et al, arxiv: ). 22

23 gg ZZ contributions Historically, computed with varying degrees of sophistication (on-shell, including decays, off-shell). Dicus, Kao, Repko (87); van der Bij, Glover (89); Matsuura et al. (91,94); Binoth et al. (08) For massless quarks circulating in the loop, again obtain compact analytic results by recycling (instead of numerical approaches previously). Analytic results ameliorate numerical stability issues. Contribution of the top quark loop is suppressed by a factor (top mass) 4 and therefore neglected. top gives a 1% difference in the gg contribution. 23

24 Importance of gg ZZ Contribution rather small (~ few %) well below threshold, but significant above (5-10%). Interference with Higgs channel also expected to be small, because of the mass resolution of this channel. 24

25 Comparing CMS and ATLAS CMS PAS ATLAS CONF Events/2 GeV/c 2 Events/2 GeV/c CMS Preliminary 2011 LEP excluded (99% C.L.) Events/10 GeV/c 2-1 s = 7 TeV L = 4.71 fb DATA Z+X ZZ 2 m H =140 GeV/c 2 m H =120 GeV/c M 4l 2 [GeV/c ] m4l [GeV/c 2 ] It would be good to see CMS data with tighter cuts and ATLAS data with looser cuts. It would be good to see lower mass values of m4l. 25

26 December 2011 analyses MH =125GeV Signal and irreducible ZZ background for CMS PAS and ATLAS CONF cuts. (Additional Z+jets background not included.) S/ B substantially the same for both experiments, although ATLAS cuts harder. 26

27 H γ γ process 27

28 One stop shopping for photons in MCFM v6.1 pp γγ including fragmentation contributions and gg γγ at NNLO pp γ+jet pp Wγ pp Zγ including fragmentation contributions including photon radiation from decay products of W (electromagnetic gauge invariance!) including photon radiation from decay products of Z. 28

29 Dealing with photons In the presence of QCD radiation (i.e. beyond LO) cross sections with photons develop additional singularities. γ γ Experimentalists impose isolation cuts on the photons singular propagator when quark and photon are collinear But isolated photon cross sections are not infrared safe 29

30 Smooth cone solution Allow soft partons, but remove collinear configurations, Frixione, hep-ph/ Parton required to be softer as it gets closer to photon. No contribution exactly at the collinear singularity. This is simple to apply to a theoretical calculation and results in a well-defined cross section. Cannot be (exactly) implemented experimentally due to finite detector resolution. 30

31 Photon fragmentation The analogous quantity to pdfs is the photon fragmentation function. γ Non-perturbative input required, but perturbative evolution. γ Using conventional isolation, only the sum of the direct and fragmentation contributions is meaningful. After isolation, the finite contribution from the fragmentation contribution is typically small. Fragmentation functions defined order-by-order; here we neglect contributions beyond LO. 31

32 Diphoton production NLO corrections straightforward to compute. At NNLO, can obtain significant contribution from diagrams with two initial-state gluons (primarily LHC). Ametller et al. (85); Dicus and Willenbrock (88) s=7 TeV LHC γ finite contribution γ 32

33 Treatment of pp γγ in MCFM MCFM includes a full NLO calculation. Fragmentation contributions are included. MCFM includes gg contributions calculated at NNLO using two loop amplitudes of Bern et al., (hep-ph/ ) MCFM combines functionality of Diphox,Jetphox and gamma2mc 33

34 Glue-glue at higher order Large contribution should be computed more accurately if possible. Next order = 2 loops: hard, but complete calculation tractable because first order is finite. Bern, De Freitas, Dixon (01); Bern, Dixon and Schmidt (02) Similar parton-level approaches: gamma2mc: Frixione-style isolation only; Schmidt Diphox: glue-glue pieces at LO only, but fragmentation treated at NLO; 2γNNLO: full NNLO calculation, Frixione-style isolation. Catani et al. (Oct. 11) Binoth et al. 34

35 Example Analysis performed using staggered photon cuts: Cuts induce a large NLO correction since photons must be produced with equal pt at LO. 35

36 Staggered cuts in perturbation theory Investigate removing staggered requirement. Replace with cuts: p γ 1 t > 40 GeV, p γ 2 T > 40 + δ GeV Cross section contains terms δ log δ. Frixione, Ridolfi (97) Resummation required for small δ (<3 GeV). 36

37 Less asymmetric cuts. The pt distributions of the second photon for the signal and background are quite different. indicates that less staggered cuts are advantageous. pt min = 25 GeV vs 37 GeV. 37

38 Reducing the stagger Modest decrease of background by increasing pt(min) of second photon from 25 to 37 GeV. At very least it would be nice to see the plots for different values of the stagger. 38

39 Conclusions MCFM offers the possibility of one stop shopping for Higgs signal and background (and much more), calculated consistently at NLO. Analytic results for gg WW through (t,b) loops available - small effect on cross section but significant for interference with Higgs diagrams. For the most sensitive WW process, signal can be increased and background decreased by performing transverse mass cut. H ZZ: Interesting to exchange style of cuts of 2 experiments H γγ: Interesting to revisit/present results with less stagger. 39

40 Preliminaries Important to retain spin correlations in W/Z decays, e.g. for analysing angle between leptons for H WW. Equivalent of radiation in decay is singly resonant diagrams, in contrast to the usual doubly-resonant set. Both WW and ZZ also receive contributions from glue-glue initiated processes. For this talk, ignore WZ (nothing new there). 40

41 Interference under search cuts Similar behavior for large Higgs masses. What is responsible for the difference at smaller mh? 41

42 Motivation Irreducible backgrounds. in various combinations: leptons, jets, missing ET; searches for the Higgs, supersymmetry,... V 1 V 1 V 2 V 2 Test of the gauge structure of the Standard Model. limits on anomalous boson couplings. 42

43 Cone problems Removing quark-photon singularities in this way would be acceptable (but only up to NLO). However, a physically meaningful prediction would also require the same cut on gluons. Enforcing such a cut would prohibit the emission of soft gluons inside the cone and be infrared-unsafe: cancellation of virtual/real singularities not complete. γ γ integration over all loop momenta soft singularities integration over all PS γ γ 43

44 Conventional approach Usually, isolation cone allows a small amount of hadronic energy inside. Okay for QCD infrared-safety, but collinear quarkphoton singularity again exposed. Singularities can be handled by usual higher-order machinery, e.g. dipole subtraction, and exposed: Just like initial-state collinear singularities are absorbed into pdfs, these can be defined away. 44